Analisa sentimentos das reclamacoes

library(tidyverse)
library(tidytext)
library(here)
library(lexiconPT)
library(plotly)
theme_set(theme_bw())

reclamacoes = read_csv(here("data/3-avaliacao-humana/reclamacoes-avaliadas-20190515.csv"))

Parsed with column specification:
cols(
  id = col_double(),
  orgao = col_character(),
  data = col_character(),
  titulo = col_character(),
  texto = col_character(),
  link = col_character(),
  `Grupo que vai avaliar` = col_double(),
  insatisfacao = col_double(),
  avaliadores = col_double(),
  range.avaliacoes = col_double()
)

#reclamacoes = reclamacoes_raw %>% 
#    mutate(
#        nome_orgao_site = orgao,
#        orgao = str_split(link, "/") %>% map_chr(~ .[[5]])
#    ) %>% 
#    filter(orgao %in% c("inss-ministerio-da-previdencia-social", #"anac-agencia-nacional-de-aviacao-civil")) %>% 
#    mutate(id = 1:n(), 
#           grupo_avaliando = id %% 6 + 1) 

O processo de estimativa sera muito baseado em https://sillasgonzaga.github.io/2017-09-23-sensacionalista-pt01/ .

data("oplexicon_v3.0")
data("sentiLex_lem_PT02")
op30 <- oplexicon_v3.0
sent <- sentiLex_lem_PT02
glimpse(op30)

Observations: 32,191
Variables: 4
$ term              <chr> "=[", "=@", "=p", "=P", "=x", "=d", "=D", ";)", ";)", ";@", ";*", ";**", ";~", ";D", ";D", ";p",…
$ type              <chr> "emot", "emot", "emot", "emot", "emot", "emot", "emot", "emot", "emot", "emot", "emot", "emot", …
$ polarity          <int> -1, -1, -1, -1, -1, 1, 1, 1, 1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 1…
$ polarity_revision <chr> "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "…

Precisamos de um dataframe onde cada observacao eh uma palavra.

palavra_a_palavra = reclamacoes %>% 
    select(id, texto) %>% 
    unnest_tokens(termo, texto)
palavra_a_palavra %>%
  select(id, termo) %>%
  head(20)

palavras_com_sentimento = palavra_a_palavra %>% 
  left_join(op30 %>% select(term, op30 = polarity), by = c("termo" = "term")) %>% 
  left_join(sent %>% select(term, sent = polarity), by = c("termo" = "term")) 

Agora, de fato, calculamos qual a polaridade acumulada (via somatorio) de cada reclamacao e salvamos em um csv.

sentimentos = palavras_com_sentimento %>% 
    group_by(id) %>%
    summarise(sentimento_op30 = sum(op30, na.rm = TRUE) *-1,
              palavras_op30 = sum(!is.na(op30)),
              sentimento_sent = sum(sent, na.rm = TRUE) *-1, 
              palavras_sent = sum(!is.na(sent)), 
              palavras = n())
sentimentos %>% 
    write_csv(here("data/5-sentimentos/sentimento.csv"))
x <- sentimentos[2]
normalized_op = ((x-min(x))/(max(x)-min(x)) * 4) + 1
y <- sentimentos[4]
normalized_sent = ((y-min(y))/(max(y)-min(y)) * 4) + 1 
reclamacoes <- reclamacoes %>% mutate(sentimentos_op = normalized_op$sentimento_op30)
reclamacoes <- reclamacoes %>% mutate(sentimentos_sent = normalized_sent$sentimento_sent)
#linearModOp <- lm(insatisfacao ~ sentimentos_op, data=reclamacoes) 
#linearModSent <- lm(insatisfacao ~ sentimentos_sent, data=reclamacoes)
#cor_op <- cor(reclamacoes$sentimentos_op, reclamacoes$insatisfacao)
#cor_sent <- cor(reclamacoes$sentimentos_sent, reclamacoes$insatisfacao)
p <- plot_ly(reclamacoes, x = ~insatisfacao, y = ~sentimentos_op, type = 'scatter')
p

No scatter mode specifed:
  Setting the mode to markers
  Read more about this attribute -> https://plot.ly/r/reference/#scatter-mode
No scatter mode specifed:
  Setting the mode to markers
  Read more about this attribute -> https://plot.ly/r/reference/#scatter-mode

q <- plot_ly(reclamacoes, x = ~insatisfacao, y = ~sentimentos_sent, type = 'scatter')
q

No scatter mode specifed:
  Setting the mode to markers
  Read more about this attribute -> https://plot.ly/r/reference/#scatter-mode
No scatter mode specifed:
  Setting the mode to markers
  Read more about this attribute -> https://plot.ly/r/reference/#scatter-mode